Abstract: In the field of deep learning, deep neural networks (DNNs) have shown good performance on classification applications. However, a DNN model is vulnerable to adversarial examples, which is ...